Timeline for What summary statistics to use with categorical or qualitative variables?
Current License: CC BY-SA 3.0
22 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 25, 2012 at 15:42 | history | edited | Michael R. Chernick | CC BY-SA 3.0 |
deleted 8 characters in body
|
| Jul 24, 2012 at 22:32 | comment | added | Michael R. Chernick | Thanks so much @Macro. Huber is a saint. i can try to be that way but it is hard. | |
| Jul 24, 2012 at 22:04 | comment | added | Macro | @Michael, I agree with you - I can't personally imagine a situation where I'd want to use a pie chart but I've yet to see a rational argument for saying that no one should ever use one. But I think, for future discussions of this type, it may suffice to get your opinion on the record and let the community decide their's - that's something whuber and chl have also encouraged me to do, as I also have an argumentative streak in me, which I'm sure you know ;) Congrats on the 10k rep, btw. I hope you'll have a look at 'tools' - you can now see deleted posts and lots of cool lists. I enjoy them. | |
| Jul 24, 2012 at 20:54 | comment | added | Michael R. Chernick | I really do think that if you are going to use the logic that misuse of a method means it should never be used when speaking of pie charts then it is not a big stretch to apply that to all of statistics. Of course I know he wouldn't except that. | |
| Jul 24, 2012 at 20:52 | comment | added | Michael R. Chernick | @whuber. I am sorry. I do not mean to be disrespectful of Peter Flom. I know him from both here and the consulting statisticians site on the ASA website. I respect his knowledge of statistics and I wasn't arguing about the content of his comments. But I did challenge his overwhelmingly strong conclusions and his evasiveness with respect to my arguments. That gets frustrating and I wanted to hear if he had a valid argument or could concede that even if pie chart can give distortion too often that is not enough justification to say they should be banned. | |
| Jul 24, 2012 at 19:46 | comment | added | whuber♦ | BTW, @Michael, I do agree with you and the arguments you are making in this thread (which I find convincing and well presented), but as a moderator I have to convey strong objections voiced by community members concerning the "tone of voice" you are adopting. Please follow the site's etiquette: stick to the subject and don't attack others. Don't even write stuff that might sound like an attack, even in jest. Of course the same admonition extends to everybody. | |
| Jul 24, 2012 at 19:43 | comment | added | whuber♦ | @Michael "A table is nearly always better than a dumb pie chart; the only worse design than a pie chart is several of them ... pie charts should never be used."--Tufte. "Data that can be shown by pie charts always can be shown by a dot chart. ... in the 1920's a battle raged on the pages of JASA about the relative merits of pie charts and divided bar charts ... both camps lose because other graphs perform far better than either divided bar charts or pie charts."--Cleveland. As you know, Cleveland is not prescriptive: this is as strong as he gets about anything. | |
| Jul 24, 2012 at 18:52 | comment | added | Michael R. Chernick | Ther probably isn't a single technique or graphical/exploratory method that hasn't been misused by someone! | |
| Jul 24, 2012 at 18:51 | comment | added | Michael R. Chernick | Gee Peter instead of just standing behind Bill Cleveland why don't you just give me a cogent argument for abolishing them? I can appreciate the argument about difficulty perceiving angle differences and about people's tendency to make poor ones but where is your counter to my argument. Do Cleveland, Tufte and other go so far as to outright state that pie chart should never be used? That seems to me would be more controversial than what I am saying. I wonder what John Tukey would have said about abandoning pie charts just because they can be misused. Maybe all of statistics should be band. | |
| Jul 24, 2012 at 18:36 | comment | added | Peter Flom | Well @MichaelChernick you are certainly free to not see it that way, but the world's experts in graphical perception disagree with you. | |
| Jul 24, 2012 at 12:12 | comment | added | chl | We have had some discussions around pie charts on this thread: Problems with pie charts. | |
| Jul 24, 2012 at 10:28 | comment | added | Michael R. Chernick | When you have 4 equal categories it may be a little easier for some to see a pie chart with 4 right angles or 45 degree angles when there are 8 categories that are equal. Bar charts may not convey this as easily. So while I agree that some people are bad a perceiving angles (other than 180 degree angles, 90 degree angles or 45 angles) and some people make bad charts, I do not see that as justification to ban pie charts. | |
| Jul 24, 2012 at 10:23 | comment | added | Michael R. Chernick | I love Tukey's stem-and-leaf diagrams. They convey more information than histograms. But i am not going to say eliminate histograms. You argue that people display poor pie charts. Then I think the answer is to teach the proper way to do do pie charts. Maybe sometimes very small categories should be aggregated. Other times there may be good reasons to show the very small categories (not to compare the very thin slices but rather to see how many small ones there are and how mcuh smaller they are compared to the big categories). | |
| Jul 24, 2012 at 10:21 | comment | added | Peter Flom | Cleveland showed, first, that people are worse at perceiving angular measurement than linear distance. Second, that changing the colors in a pie chart changed people's perceptions of the size of the slices. Third, that rotating the pie chart changed people's perceptions of the size of the slices. Fourth that people had trouble ordering the slices from largest to smallest unless they were very different sized. Cleveland dot plots avoid all these. | |
| Jul 24, 2012 at 10:03 | comment | added | Michael R. Chernick | @Peter Flom I am familiar with some of Bill Cleveland's work and books but not with all this negativity about pie charts. Can you give me a good argument against them? We all know that there are many presentations of bad graphics and there is an art to presenting good graphics. Tufte has been excellent at showing the bad and demonstrating ways to improve it. Huff's classic "How to Lie with Statistics" shows the many unethical ways subtle or not so subtle manipulation cn change the story that a graph conveys. But do we stop showing linear trends because we canlie by doing scla manipulations? | |
| Jul 23, 2012 at 22:03 | comment | added | Peter Flom | Pie charts distort the data; with large numbers of categories they are unreadable. People are bad at perceiving angles. See books by William S. Cleveland (he also has a web site). | |
| Jul 23, 2012 at 20:21 | comment | added | Michael R. Chernick | @PeterFlom. I don't know. What is your big objection to pie charts? They never did anything bad to me. | |
| Jul 23, 2012 at 11:31 | comment | added | Peter Flom | They are certainly popular! But I think it's part of our responsibility, as experts in the field, to make pie charts less popular. | |
| Jul 23, 2012 at 11:26 | history | edited | Michael R. Chernick | CC BY-SA 3.0 |
edited body; added 20 characters in body
|
| Jul 23, 2012 at 11:25 | comment | added | Michael R. Chernick | @PeterFlom My point was not to list all the possiblr graphical procedures for summarizing qualitative data. I really want to emphasize that it is really proportion that can be compared and the way the proportions are distributed across the categories. For visually recognizing differences in proportions I think bar charts are easier to visualize than pie charts but they are just two popular ways to summarize categorical data. I don't want to say they are the best as I am not familiar with all the available methods. | |
| Jul 23, 2012 at 11:18 | comment | added | Peter Flom | I agree about charts, but not about your recommendation of pie charts. Cleveland dot plots are much superior. I have a presentation on this coming up at NESUG: Graphics for univariate data: Pie is delicious but not nutritious | |
| Jul 23, 2012 at 11:09 | history | answered | Michael R. Chernick | CC BY-SA 3.0 |